Cognitive skills, developed when individuals are young, are key to later-life outcomes, including labor market outcomes and health. Not surprisingly, researchers in the behavioral and social sciences have devoted a great deal of scholarly effort to this topic, conducting hundreds of studies designed to sort out causal mechanisms involved in the relationship between cognitive development and later-life outcome. The ultimate goal of this research agenda is to provide guidance to a vitally important policy arena-society's efforts to improve cognitive development among young people. Many of the papers in the literature use the National Longitudinal Studies of Youth (NLSY), which are arguably the most important individual-level longitudinal data in the U.S. available for understanding the antecedents of individuals' lifetime outcomes. These data are valuable for this purpose because they include data on cognitive ability, measured when individuals are young, and also have data on such outcomes as completed education, occupational choice, income, employment and unemployment, location of residence, criminal activity, marriage and fertility, criminality, and more. This application seeks funding for a project that will substantially improve data provision in the National Longitudinal Surveys of Youth (NLSY) for the purpose of studying the relationship between cognitive ability and later life outcomes. Further, we seek to develop new statistical methods that will allow researchers to make appropriate use of the new data. As we demonstrate in our Research Strategy, there are good reasons to be concerned about statistical practice in the current literature. In particular, researchers typically take estimates of cognitive ability (e.g., the AFQT score in the NLSY79 or similar test scores supplied in other data sets) and use them as regular data points for further statistical analysis. The problem here is that cognitive ability is a latent construct; test scores are simply a noisy measure of that construct. Standard practice can thus lead to badly biased estimates and, therefore, to serious misunderstandings about the relationships under study. We argue that a better way to proceed is with models that take proper account of the stochastic nature inherent in the estimation of cognitive ability, and incorporate that variability when estimating regressions of interest. The statistical methods we introduce are data hungry; they require that researchers have the full array of item responses for instruments used to assess cognitive ability. Thus we propose, wherever technically and legally feasible, to provide such data as part of the NLSY public use releases. We propose further to develop statistical models that use the data appropriately, and we intend to disseminate that knowledge to the broader research community.